University of Wisconsin–Madison Institute for Research on Poverty

Volume 24

Number 3 Focus Fall-Winter 2006

Does temporary agency employment offer a way After welfare reform: You choose your child out of poverty? 1 over the job 25 Race and poverty: Divergent fortunes of Observation in poverty research 29 America’s children? 8 New perspectives on the social and economic Welfare reform as a failed political strategy: contexts of HIV/AIDS infection 33 Evidence and explanations for the stability of public opinion 17 ISSN: 0195–5705

Does temporary agency employment offer a way out of poverty?

Is temporary agency employment a good strategy for In the research summarized here, David Autor and Susan helping the poor and disadvantaged escape from their Houseman took advantage of a unique, multiyear policy poverty? Resolving this question is no small matter. Al- “experiment”—the quotation marks are used to empha- though the temporary help sector accounts for less than 3 size that the experiment was created by circumstance percent of average U.S. daily employment, it employs a rather than by deliberate intention—to provide direct evi- disproportionate share of minority and low-skilled work- dence on the longer-term labor market effects of tempo- ers. And state administrative data show that between 15 and 40 percent of former welfare recipients who obtained employment after the 1996 welfare reforms took jobs in The research summarized here is reported at the temporary help sector.1 length in David H. Autor and Susan N. House- man, “Temporary Agency Employment as a Way Several recent studies present evidence that low-skilled out of Poverty?” which appears as a chapter in workers who take temporary agency jobs are more likely Working and Poor: How Economic and Policy than other low-skilled workers (particularly the unem- Changes Are Affecting Low-Income Workers, ed. ployed) to later transition into stable employment.2 These R. Blank, S. Danziger, and R. Schoeni (New York: findings advance the intriguing possibility that it might be Russell Sage Foundation, December 2006; see beneficial to encourage low-skilled workers to take tempo- box p. 7). rary agency jobs as a ‘stepping stone’ into the labor market. David H. Autor is Associate Professor of Econom- However, it is not entirely clear that the positive correlation ics at the Massachusetts Institute of Technology between obtaining temporary help agency employment and and Susan N. Houseman is Senior Economist at finding stable employment should be used to guide policy. the W. E. Upjohn Institute for Employment Re- In practice, it is difficult to disentangle the respective roles search. A previous Focus article by these authors of worker characteristics and job characteristics in engen- is “The Role of Temporary Employment Agencies 3 dering success in the labor market. Thus, to establish the in Welfare to Work: Part of the Problem or Part of true effect of temporary help agency employment on labor the Solution?” Focus 22, no. 1 (Special Issue, market advancement, something akin to a randomized-con- 2002): 63–70. trolled experiment would be ideal. rary sector employment. In particular, they used quasi- week in program activities or job search until they were experimental data from a Michigan welfare-to-work pro- successful. If they failed to meet their obligations, they gram, “Work First.” were terminated from Work First and faced sanctions including, ultimately, an end to TANF benefits. What might be the consequences of temporary help em- ployment for poor and unskilled people? Some argue that The Work First program structure and set of services temporary help agencies, which face lower employee were largely standardized among contractors. Contrac- screening and hiring costs than do direct-hire employers, tors typically spent one week providing new participants may choose to hire individuals who otherwise would have with basic job search skills and strategies—skills assess- difficulty finding any job at all. Such workers will thereby ment and employability planning, interviewing and pre- more quickly find work; they may begin to develop the sentation, and life-skills training. Except for “tech-prep” skills and labor market contacts that ultimately lead to courses, which quickly reviewed skills that might be more stable and better-paying jobs. But a case can also be tested, they provided little in the way of remedial, voca- made that temporary help agencies provide few opportu- tional, or computer skills training. After the first week, nities or incentives for workers to develop skills and participants were expected to look for jobs full time. networks, instead offering a series of unstable, low- Virtually all contractors provided individual job search skilled, low-paid, dead-end jobs.4 Worse, spells of tem- assistance and referrals to specific employers. They ac- porary employment may divert individuals from more companied participants to job fairs, brought employers productive searches for stable, longer-term employment. on site to recruit participants, and sponsored activities such as job clubs. Participants were required to take any Both these scenarios could be correct. In some situations job offered them that paid the federal minimum wage and companies may use temporary agencies to screen individu- satisfied work hours requirements. als for permanent jobs with good pay, benefits, and career ladders. For social welfare policy, however, what matters is Once participants found suitable jobs, contractors had to the effect of temporary employment on the low-skilled follow up with them and their employers until the partici- workers targeted by government programs. The empirical pant earned enough to terminate the TANF case or was analyses Autor and Houseman report here suggest that short- terminated from the program (usually for noncompli- term benefits from temporary agency employment do not ance).6 The Work First provider’s contract with the city persist, and that over the longer term—a year or more— was for one year, and the contractor’s performance was participants placed in temporary jobs are no better off than if evaluated by the fraction of participants who got jobs and they had received no job placement at all. the percentage who held those jobs for 90 days.

To examine the effects of temporary agency employment The “Work First” program in Michigan on the low-skilled workers in the Work First program Autor and Houseman drew on two types of evidence. The To meet the requirements of federal welfare legislation, first is a telephone survey of Work First contractors oper- most states, including Michigan, implemented a “work ating in Detroit, conducted from fall 2004 to spring 2005. first” strategy after 1996. Applicants to Temporary Assis- The second comprises state administrative and earnings tance for Needy Families (TANF) who did not already data on all participants who entered the program from meet mandatory work requirements were obliged to par- 1999 to 2003. The Work First data included demographic ticipate in welfare-to-work programs, called Work First information on Work First participants and detailed in- in Michigan, to help them find employment. Michigan formation on the jobs they found. The state Unemploy- required most TANF recipients to work 40 hours a week ment Insurance records enabled the researchers to track to remain eligible for assistance. the earnings and labor market histories of participants both during the program and earlier. In all, they examined In Michigan, the state Department of Human Services some 36,000 Work First spells of activity.7 Their data determined welfare eligibility and TANF benefits, but covered 9 geographic districts in which 25 contractors local agencies administered welfare-to-work programs. operated programs. Autor and Houseman studied the Work First program operating in Detroit, which was divided into geographic The Work First participants in the sample were predomi- districts; within each geographic district, one to three nantly female (94 percent) and black (97 percent). Their nonprofit private or public agencies contracted to provide average age was 30, and over a third had not completed employment services. When multiple contractors pro- high school. Slightly under half of the women found jobs vided Work First services within a district, they took while in Work First. Among participants who found jobs, turns enrolling applicants. In these districts the distribu- about 20 percent held jobs through temporary help agen- tion of participants among contractors was thus equiva- cies while the remaining 80 percent obtained direct-hire lent to random assignment.5 At the time this study was jobs. Relative to those who obtained temporary help or undertaken, Work First participants were required to treat direct hire-jobs, those who did not obtain any job were the program as if it were a job and to spend 40 hours a more likely to have dropped out of high school, had 2 Never

Sporadically

Every Few Months

Monthly

Weekly

0% 5% 10% 15% 20% 25% 30% 35% 40% Percentage of Contractors Reporting Invite temp agencies on site Refer participants to temp agency jobs

Figure 1. Contractors’ contacts with temporary work agencies. worked fewer quarters before entering the program, and ence of the contractors on whether or not clients obtain had lower prior earnings. temporary help jobs (even allowing for some inflation in their reporting).

The variability in contacts and placement reflects large dif- Contractors and temporary employment ferences among contractors’ views of the value of temporary agencies: Findings from the survey agency placement. Of the 21 contractors, 13 claimed to take a neutral stance on the issue, five discouraged temp jobs, and In the telephone survey, Autor and Houseman inter- three reported that they encouraged such jobs. Over three- viewed the heads of 21 contractor organizations that had quarters of contractors thought that direct-hire jobs lasted operated programs when participants in the study were longer and were more stable, and more than half thought enrolled in Work First. They found almost no differences they paid as well or about the same as temporary agency among contractors’ resources or program services that jobs. All who discouraged temporary agency jobs, and most might affect what happened to participants. They sought of those who were neutral, believed that temporary agency to determine if contractors played a significant role in jobs generally did not lead to permanent positions. Even two whether participants found jobs and, by implication, in of the three who encouraged temporary agency employment where they found them. Contractors themselves estimated said they did so only in cases when the position was explic- that they were directly involved in around 75 percent or itly “temp-to-hire” or when direct-hire options were poor. more of job placements, through referrals, on-site visits, There was little disagreement about the benefits of using and the like. Most strikingly, contractors varied substan- temporary help agencies for temp-to-hire placements. But tially in the amount of exposure to temporary agency interviewees noted that companies screening for permanent work that they provided to participants and in the propor- employees through an agency were usually looking for more tion they placed in such jobs (Figure 1). Among the 17 skilled workers. Such workers would be relatively easy to contractors who provided an estimate, the median place- place in either direct-hire or temporary agency positions. ment rate of clients in temporary agency jobs was 15 percent; three contractors reported that 5 percent or fewer Work First providers faced real difficulties in placing of job placements were with temporary agencies, and participants with very low skills, little or no work experi- three others reported that agency placement accounted ence, or poor work ethics—individuals for whom direct- for a quarter, a third, and three-quarters of all such place- hire positions are often scant. Contractors were also un- ments. Placement with a temporary agency was closely der considerable pressure to increase their job placement correlated with the frequency of agency visits and refer- rates. Given such constraints, over half the contractors rals—a finding that tends to confirm the significant influ- thought that temp agency jobs were easier for those with

3 weak skills or experience to obtain; indeed, 43 percent effect. It is not possible simply to compare subsequent viewed such jobs as the only realistic alternative to unem- employment and earnings among Work First participants ployment for some participants. These contractors be- who obtain direct-hire jobs, temporary jobs, or no job lieved that some clients were simply not ready to hold a while in the program, because these participants have permanent job, and that agencies could play a role in very different characteristics. Those placed in direct-hire giving them work experience and an understanding of and temp jobs had higher earnings during the next two employers’ expectations. Temporary employment al- years than those not placed in a job. But they also had lowed participants to sample different jobs, make con- significantly higher levels of education and earnings be- tacts with many different employers and, if they did not fore they entered Work First. And because participants’ like a particular job, request reassignment to avoid the observable characteristics—such as education and prior stigma of quitting. Flexibility in work scheduling was earnings—differed, it is highly likely that their unmea- also widely cited as an advantage of temp work—almost sured characteristics, such as motivation and employment two-thirds of contractors thought that temporary agency barriers, also differed. jobs were better at accommodating participants’ needs. In the Autor and Houseman study, the quasi-experimental Contractors who worked with temporary help agencies nature of the Detroit Work First program was key to tended to stick to one or more agencies with which they identifying the true effects of placement into a temporary had developed specific relationships and clear under- help job, a direct-hire job, or no job on subsequent em- standings about the duration of job assignments and the ployment, earnings, and welfare outcomes. As noted prospects for permanence. They avoided agencies that above, Work First participants were, in effect, randomly offered very short-term employment, for example, assign- assigned to contractors, and so contractors operating in ments with day-labor agencies. the same district had comparable populations (more pre- cisely, identical populations in a statistical sense). These The long-term consequences of temporary agency contractors had different job placement practices, how- placements ever, but otherwise provided virtually identical services. The most important question for welfare-to-work policy Estimating the longer-run effects of temporary sector is whether temporary agency placements foster stable, employment longer-term employment, ideally at wages that can sup- port a family. On this, contractors did not agree. More Using this quasi-experimental design, Autor and Houseman than half of the contractors believed that temp placements estimated how the Work First participants would fare over helped participants develop confidence, but there were time if they were placed in a direct-hire or a temporary offsetting effects. A majority thought that temp place- agency job, relative to no job at all.8 Both direct-hire and ments allowed participants to avoid making a serious temporary placement significantly increased participants’ commitment to employment. For one thing, a contractor employment and earnings (relative to no job placement) commented, a temporary agency job “allows a person to over the quarter following program entry. Direct-hire place- stay compliant with Work First requirements, so they ments continued to do so for two years after program entry. can’t be terminated from the program. But unless the job In contrast, the early positive effects of temporary agency is temp-to-hire, they will end up back in the program. It job placements were short-lived. Two to eight quarters fol- could be six months, it could be a year later, but they will lowing their entry into the program, the employment and end up back in the program, in the same place. They earnings of those initially placed in temporary agency posi- won’t have made any advances.” tions showed no increase at all in employment or earnings, relative to those with no placement. And those in temp The divergent policies and practices of contractors who positions had significantly lower employment rates and la- provided services to the same Work First population and bor earnings than those in direct-hire jobs. operated in the same labor market resulted in different rates of placement into temporary-agency jobs, direct- To examine whether temporary agency placements hire jobs, or no job. Coupled with the random assignment helped participants leave welfare and escape poverty, of participants among contractors, these different prac- Autor and Houseman used several measures. First, they tices enabled Autor and Houseman to identify the longer- determined the proportion of participants who left Work term labor market effects of temporary agency place- First because they had obtained a job that paid enough to ments, in particular, their effect on the probability of close their TANF cases. This is the immediate and ex- leaving welfare and escaping poverty. plicit goal of the Work First program, so that a case closed because of earnings is considered to be an indica- tor of the success of the program. Overall, about 18 percent of Work First spells in the sample ended because What the administrative data say of earnings. About 38 percent of participants with a di- rect-hire job and 33 percent of those with a temporary A key challenge, as noted above, for any empirical inves- agency job achieved earnings levels sufficient to close tigation of this sort is establishing a chain of cause and their TANF cases during their Work First spell. Using the 4 Table 1 Participants with Earnings Sufficient to Escape Welfare or Poverty, during the Two Years after Their Work First Orientation

Welfare Threshold, Welfare Threshold, Poverty Threshold, Poverty Threshold, Family of 3 Family of 4 Family of 3 Family of 4

Quarter 1 (N = 36,105) No job 6.7 % 5.4 % 3.4 % 2.1 % Direct-hire job 21.3 17.0 9.9 5.2 Temp job 21.1 16.4 9.7 5.2

Quarters 1-4 (N = 36,105) No job 9.1 7.1 4.2 2.2 Direct-hire job 22.1 17.5 10.0 5.3 Temp job 21.3 17.1 9.5 5.4

Quarters 5-8 (N = 25,118) No job 12.2 10.1 8.8 5.1 Direct-hire job 19.8 16.5 15.0 8.5 Temp job 19.3 16.0 14.5 8.9

Note: Panels labeled Quarter 1 and Quarters 1–4 include Work First participants entering from the third quarter of 1999 through the second quarter of 2003. Participants in the last panel (Quarters 5–8) entered from the third quarter of 1999 through the second quarter of 2002. quasi-random-assignment design, the authors find that porary employment showed different short- and medium- direct-hire job placements significantly increased the term effects. They indicated that in the near term (the first probability of a successful case closure based on earn- quarter after program entry), both direct-hire and tempo- ings, but temporary agency employment did not.9 rary employment significantly raised the probability that participants’ earnings would exceed the welfare or pov- The case-closure measure was important to contractors erty thresholds. The picture was distinctly different when because they were evaluated by it, but from a broader the time horizon was extended. Direct-hire jobs signifi- policy perspective it is potentially flawed. Participants cantly increased the probability that earnings would ex- who do not find a job while in the program may find ceed the income level necessary for a family of three or employment on their own and leave welfare, but are not four to remain off welfare for the two years after entry. counted as “successes” by this measure. And individuals For example, the earnings of a woman placed in a direct- who are terminated because of earnings might lose the job hire job were more likely to exceed the welfare threshold and be back on welfare and in Work First in a relatively for a family of four by 5.9 percentage points in the first short time. To overcome these difficulties, Autor and year and by 10.8 percentage points in the second year. In Houseman examined whether participants earned enough contrast, temporary agency placements did not appear to over a longer time to end their dependence on welfare and help participants stay off welfare. The effects of such escape poverty. Because the data did not allow them to placements on the probability of earning above the wel- compute individual welfare and poverty thresholds for fare threshold were insignificant and even negative the sample, they selected several thresholds as measures (though very small). of success. They used both welfare and poverty thresh- olds because the income threshold for welfare benefits is For the higher poverty thresholds, neither temporary considerably below the poverty level for any given family agency nor direct-hire placements helped participants es- size. In Michigan, earnings needed to cut welfare benefits cape poverty over the one-to-two-year period following for a family of three with two dependent children, for their entry into the program. And relative to direct-hire example, were $9,504 in 2003, only 64 percent of the placement and even to no placement at all, the effects of poverty level. temporary agency placement on the probability that par- ticipants earned above the poverty threshold for a family Relatively few participants attained earnings even above of four in the year after entry were significantly negative, the lowest threshold, as Table 1 shows. In the first four though again small. quarters after program entry, for example, just over 20 percent of those with either type of job earned above the welfare threshold for a family of three. In the first year Should employment programs use temporary after they entered the program, around 5 percent of those agency placement? working earned above the poverty threshold for a family of four with three dependent children. These analyses found no evidence that suggests that em- ployment programs would gain by increasing the use of Regression analyses which took advantage of the agency temporary employment agencies as labor market interme- variation in placement rates to identify the effect of tem- diaries for low-skilled workers. Although such agencies

5 did raise the probability that Work First participants modate these barriers will be more beneficial in the long earned above the welfare and poverty thresholds over run. And to that end, an incentive structure for contrac- very short time horizons, these effects quickly dissipated. tors that stresses longer-term outcomes might reduce wel- Over one to two years, temporary agency placements did fare dependence and poverty levels by easing the pres- not increase the chances that participants would earn sures on program providers to move participants quickly enough to leave welfare and escape poverty, and by some into any job available, whether or not these jobs are likely measures they even reduced these chances. Raising the to offer longer-term benefits to participants. number of direct-hire placements is likely to be a much more effective means for job assistance programs to re- duce welfare dependency over the short and the long 1In 1999, African Americans were overrepresented in the ranks of term. temp workers by 86 percent, Hispanics by 31 percent, and high-school dropouts by 59 percent; college graduates, by contrast, were These results do not imply that temporary agency jobs underrepresented by 47 percent (M. DiNatale, “Characteristics and Preference for Alternative Work Arrangements, 1999,” Monthly La- never improve long-term outcomes for participants. Cer- bor Review 124, no. 3 [2001]: 28–49). On temporary sector employ- tainly such cases occur on a regular basis. What the Autor ment among welfare recipients, see, e.g., C. Heinrich, P. Mueser, and and Houseman analysis shows, however, is that, on aver- K. Troske, “Welfare to Temporary Work: Implications for Labor Mar- age, the benefits of temporary help job placements in- ket Outcomes,” Review of Economics and Statistics 87, no. 1 (2005): duced by job assistance programs are nil. Thus, the Autor 154–73 (on North Carolina and Missouri); M. Cancian, R. Haveman, T. Kaplan, and B. Wolfe, Post-Exit Earnings and Benefit Receipt and Houseman analysis directly addresses the germane among Those Who Left AFDC in Wisconsin, IRP Special Report 75, policy question “Can job assistance programs improve University of Wisconsin–Madison, 1999. labor market outcomes for participants by placing more 2Heinrich and colleagues, “Welfare to Temporary Work”; M. Ferber clients in temporary agency positions?” Their analysis and J. Waldfogel, “The Long-Term Consequences of Nontraditional suggests the answer is generally “No.” Employment,” Monthly Labor Review 121, no. 5 (1998): 3–12; and M. Corcoran and J. Chen, “Temporary Employment and Welfare to Why should this be the case? The analysis suggests some Work,” unpublished paper, University of Michigan, 2004. plausible explanations. Contractors reported that tempo- 3H. Holzer, “Encouraging Job Advancement among Low-Wage Work- rary help jobs were more plentiful than direct-hire jobs, ers: A New Approach,” Welfare and Beyond Policy Brief no. 30, particularly for those with weak skills and experience, but Brookings Institution, Washington, DC, 2004; J. Lane, K. Mikelson, P. Sharkey, and D. Wissoker, “Pathways to Work for Low-Income generally did not lead to permanent positions. According Workers: The Effect of Work in the Temporary Help Industry,” Jour- to contractors, some temporary agency jobs could pro- nal of Policy Analysis and Management 22, no. 4 (2003): 581–98. vide useful entrees into permanent placements with good 4For discussions of the various hypotheses, see D. Autor, “Why Do employers. But these temp-to-hire jobs generally re- Temporary Help Firms Provide Free General Skills Training?” Quar- quired skills and experience that their clients typically terly Journal of Economics 116, no. 4 (2001): 1409–48; S. Houseman, did not possess. The researchers’ estimates indicate that “Why Employers Use Flexible Staffing Arrangements: Evidence from temporary agency placements subsequently resulted in an Establishment Survey,” Industrial and Labor Relations Review 55, increased earnings in the temporary help sector but re- no. 1 (2001): 149–70; and H. Jorgenson and H. Riemer, “Permatemps: Young Temp Workers as Permanent Second Class Employees,” duced earnings in direct-hire jobs. This implies that many American Prospect 11, no. 18 (2000): 38–40. of the participants who obtained temporary help jobs 5For an analysis of the random nature of the assignment, see D. Autor through job assistance programs would have been likely and S. Houseman, “Do Temporary Help Jobs Improve Labor Market to obtain direct-hire work on their own (perhaps after a Outcomes for Low-Skilled Workers? Evidence from Random Assign- longer period of job search). The analysis also finds some ments,” Working Paper 05-02, National Poverty Center, University of evidence that participants placed in temporary agency Michigan, Ann Arbor, September 2005. jobs are more likely than other participants to continue 6The median Work First spell in our sample was slightly under 3 working in the temporary help sector and more vulnerable months, and 96 percent of participants were terminated from the to frequent spells of non-employment and shorter periods program in less than a year. of work. Indeed, participants placed in temporary help 7Many participants engaged in more than one Work First spell—36 jobs during the first Work First spell were significantly percent of the sample reentered Work First within a year of their first more likely to be back on welfare within two years than entry and over half were back within two years. were participants placed in direct-hire jobs. 8We estimated both OLS and two-stage least squares models; the latter permit a causal interpretation. Some of the contractors interviewed held the view that a 9Our two-stage least squares models indicate that direct-hire place- certain segment of the temporary agency market accom- ment significantly increased (by nearly 25 percent) the probability modated individuals with very weak skills, experience, that a participant would reach this earnings level. Temporary agency employment, although still positive (11.5 percent), is insignificant. and motivation to work by providing them with jobs that required few skills and no long-term commitment. Yet these are precisely the qualities that tend to keep such individuals in poverty. It is plausible that job placements that might help individuals overcome rather than accom-

6 Working and Poor: How Economic and Policy Changes Are Affecting Low-Wage Workers Rebecca M. Blank, Sheldon H. Danziger, and Robert F. Schoeni, editors Over the last three decades, large-scale economic developments, such as technological change, the decline in unionization, and changing skill requirements, have exacted their biggest toll on low-wage workers. These workers often possess few marketable skills and few resources with which to support themselves during periods of economic transition. In Working and Poor various scholars examine how economic and policy changes over the last twenty-five years have affected the well-being of low-wage workers and their families. Introduction: Rebecca M. Blank/Sheldon H. Danziger/Robert F. Schoeni Part I. What is Changing (and Why) in the Labor Market for Low-Skilled Workers? 1. Exploring Gender Differences in Employment and Wage Trends Among Less-Skilled Workers Rebecca M. Blank/Heidi Shierholz 2. Wage Trends among Disadvantaged Minorities George J. Borjas Part II. How do Economic Trends Affect Less-Skilled Workers? 3. The Macroeconomy and Determinants of the Earnings of Less-Skilled Workers Robert Hall 4. The Impact of Technological Change on Low-Wage Workers: A Review David Card/John DiNardo 5. The Chaging Pattern of Wage Growth for Low-Skilled Workers Eric French/Bhashkar Mazumder/Christopher Taber Part III. How do Macroeconomic Changes Influence Well-Being Measures Beyond Income? 6. The Level and Composition of Consumption Over the Business Cycle: The Role of “Quasi-fixed” Expenditures Kerwin Kofi Charles/Melvin Stephens Jr. 7. Recent Trends in Resource Sharing among the Poor Steven J. Haider/Kathleen McGarry 8. Economic Conditions and Children’s Living Arrangements Rebecca A. London/Robert W. Fairlie Part IV. How do Policy Changes Interact with the Economy and Economic Well-being? 9. How do Tax Policies Affect Low-income Workers? Kevin A. Hassett/Anne Moore 10. State Spending on Social Assistance Programs Over the Business Cycle Therese J. McGuire/David F. Merriman 11. Temporary Agency Employment as a Way out of Poverty? David Autor/Susan Houseman 12. Child Support and the Economy Maria Cancian/Daniel R. Meyer 13. Unemployment Insurance over the Business Cycle: Does it Meet the Needs of Less-Skilled Workers? Phillip B. Levine 14. How is Health Insurance Affected by the Economy? Public and Private Coverage among Low-skilled Adults in the 1990s Helen Levy

416 pp., hardcover, 2006, $49.95 Russell Sage Foundation, 112 64th Street, New York, NY 10021 Order online (http://www.russellsage.org) or call 1-212-750-6000

7